基於使用者密度之多速率及多基地台WIFI網路選擇

Abstract

本論文討論及分析效能時以throughput為主，參考目前的無線網路環境，使用IEEE 802.11標準。在多天線的支援下,實體層資料傳輸能提供多種傳輸速率，其實質系統throughput卻受限於媒介存取控制(Media Access Control, MAC)層機制所採用的分散式協調功能(Distribution Coordination Function, DCF)。由於在同一AP中，此方案保障所有裝置長期存取通道次數機會,使得低速率裝置佔用較長的通道時間，降低各個裝置通到存取機會。在多AP和多速率傳輸的部分，現今的方法是選擇各個可存取AP的RSSI (Received Signal Strength Indicator)裡最大值以達到最高傳輸速率之AP來進行傳輸，但是卻使眾多裝置連上同一AP形成過載的結果，最終造成使用者所實得throughput降低。因此我們在各個AP中使用分群的方式將不同傳輸速率的使用者分時傳輸,避免高速率和低速率在同一AP傳輸中造成throughput下降。並且我們針對使用者角度提出基於使用者密度網路選擇法，透過無線網路的模擬與分析設計出不同傳輸速率的方程式，透過方程式判斷使用者可存取的AP中擁有最大throughput，對該AP進行存取。最終我們和現有的高速率導向網路選擇方法比較,發現在不同使用者分佈下,使用者密度網路選擇法的throughput皆高於現有的高速率導向網路選擇法，在最好的情況下使用者密度網路選擇法比現有的高速率導向網路選擇法高出400%的效能。

In this thesis, we discuss and analysis performance based on throughput. According to IEEE 802.11 standard, physical layer provide multiple data rate supported by multiple antennas. However, the actual throughput is limited with MAC (Media Access Control) layer mechanism that adapting DCF (Distribution Coordination Function). In DCF, each client has the same probability to access the channel in the AP in long time transmission so that low rate clients occupy longer time to access channel. As a result, the access time of clients is reduced and the throughput is lower. On the other hand, in multi-rate multi-AP system, client selects the highest RSSI (Received Signal Strength Indicator) from candidate APs to achieve the greatest data rate for existing method. However, this approach leads to overloading of particular AP so that actual client throughput is lower. Ergo, we use grouping to separate clients based on different data rate in groups to avoid lower throughput due to contention together from different rate in single AP. For multi-AP, We propose PD (Population Density) based network selection from viewpoint of client. In PD based network selection, we fit functions based on different data rate according to simulation and analysis. The best client throughput of all connections can be determined by functions and access the AP which provide maximal client throughput. After designing, we compare performance of PD based network selection with those of exiting HR (high rate) based network selection. PD based is always have better performance than HR based network selection. Also, the performance improvement of PD based network selection is 400% in the best case.